Holonic Intelligence: A Paradigm Shift

Transcription

Holonic Intelligence:
A Paradigm Shift
William A. Gruver
Intelligent Robotics Corporation
Simon Fraser University
Workshop on Intelligent Systems – Festschrift for Dr. Richard Volz
Texas A&M University, College Station, TX
April 10, 2010
Outline
Background
Motivation for distributed intelligence
Comparison with centralized intelligence
How to achieve distributed intelligence
Technologies

Multi-agent and holonic systems
Cooperation, collaboration, coordination
Holonic intelligence system architecture
Holonic intelligence network
Applications
Manufacturing automation, decision support
Energy management, smart grid
Smart home, digital services
WAG/10apr10
Intelligent Robotics Corporation • Simon Fraser University
Texas A&M 2
Why is a paradigm change needed?
Autonomous systems and robotic technologies are
becoming pervasive
Unmanned system capabilities are present in many
space and combat systems
Service robots are being developed for widespread
use and varied applications
.
WAG/10apr10
Texas A&M 3
Why is a paradigm change needed?
System of Systems (SoS)
Availability of feature rich sensors, actuators, and
controllers
Increasing trend to network appliances and combine
their controls and key functions
WAG/10apr10
Texas A&M 4
Distributed Intelligence in Nature
Each ant has simple intelligence
− distributed intelligence
Communicates with other ants
− distributed communications
Uses pheromones to communicate
Key decisions
− food found, follow food found pheromone
− food not found, find food elsewhere
− return to colony
WAG/10apr10
Texas A&M 55
Centralized systems are everywhere …
WAG/10apr10
Texas A&M 6
Control / Knowledge Imbalance
Control
Knowledge
WAG/10apr10
Texas A&M 7
Contradictory Nature of Technology
Internet:
Designed for peer-to-peer
communications but the Web
has a client/server architecture
Wireless Communications:
Hierarchical infrastructure, but
increasing demand for peer-topeer applications
Information Systems:
Data, knowledge, information
are concentrated but activities
are distributed
WAG/10apr10
Texas A&M 8
Hierarchical Organizational Model
“The work of every workman is fully planned out by the
management at least one day in advance, and each man
receives in most cases complete written instructions,
describing in detail the task which he is to accomplish, as well
as the means to be used in doing the work. … This task
specifies not only what is to be done, but how it is to be done
and the exact time allowed for doing it. … Scientific
management consists very largely in preparing for and
carrying out these tasks.”
Frederick Taylor, Principles of Scientific Management, 1911
WAG/10apr10
Texas A&M 9
Centralized Systems
WAG/10apr10
Texas A&M 10
Disadvantages of Centralized Systems
Scalability
– Servers have finite storage
and finite processing
Robustness
– Servers may not be able to
respond to clients
Security
– Additional security needed
to prevent unauthorized
access
Communications
– Limited communication
paths
WAG/10apr10
Texas A&M 11
Distributed System
each node contains a unique subset of the system
information
each node processes a unique subset of the system
tasks
WAG/10apr10
Texas A&M 12
But where are we today?
Application
Application
Presentation
Presentation
Session
Session
Transport
Transport
Network
Network
Data Link
Data Link
(MAC LLC Layers)
(MAC LLC Layers)
Physical
Physical
OSI was designed for point-to-point connections in client/server
applications.
WAG/10apr10
Texas A&M 14
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
RPC
Pipes
CORBA
ObjectOriented
Messaging
Sockets
DCOM
RMI
.NET
FIPA ACL
Agent
Messaging
ICM
April
KQML
Solutions have been developed for many different kinds of system
architectures, further complicating the development of distributed systems.
WAG/10apr10
Texas A&M 15
Jason
There are many environments for developing distributed systems, but they
often complicate the problem instead of simplifying it.
WAG/10apr10
Texas A&M 16
TCP/IP Model
OSI Model
HTTP, SMTP, SNMP, FTP, Telnet, SSH, Scp, NFS, RTSP
XDR, ASN.1, SMB, AFP
TLS, SSH, ISO 8327 / CCITT X.225, RPC, NetBIOS, ASP
TCP, UDP, RTP, SCTP, SPX, ATP
IP, ICMP, IGMP, X.25, CLNP, ARP, RARP, BGP, OSPF, RIP, IPX, DDP
Ethernet, Token ring, PPP, HDLC, Frame relay, ISDN, ATM, 802.11 WiFi, FDDI
wire, radio, fiber optic
Furthermore, there are too many protocols …
WAG/10apr10
Texas A&M 17
HTTP, SMTP, SNMP, FTP, Telnet, SSH, Scp, NFS, RTSP
XDR, ASN.1, SMB, AFP
TLS, SSH, ISO 8327 / CCITT X.225, RPC, NetBIOS, ASP
TCP, UDP, RTP, SCTP, SPX, ATP
IP, ICMP, IGMP, X.25, CLNP, ARP, RARP, BGP, OSPF, RIP, IPX, DDP
Ethernet, Token ring, PPP, HDLC, Frame relay, ISDN, ATM, 802.11 WiFi, FDDI
wire, radio, fiber optic
… that require more programming at the application level
WAG/10apr10
Texas A&M 18
Overloaded Application Domain
Because of limited intelligence in the lower layers of the OSI model, higher layers
are needed to perform networking functions.
WAG/10apr10
Texas A&M 19
What’s the solution?
Application
Application
Presentation
Presentation
Session
Session
Transport
Transport
Network
Network
Multiple MAC Layer Management
Data Link
(MAC LLC Layers)
Physical
Data Link
Data Link
Data Link
(MAC LLC Layers)
(MAC LLC Layers)
(MAC LLC Layers)
Physical
Physical
Physical
First, we need multiple simultaneous connections and multi-hop services
WAG/10apr10
Texas A&M 21
Next, we need intelligent
multi-agent systems that can
handle network services
WAG/10apr10
Texas A&M 22
… so that applications are only concerned with
application specific services and how to interoperate
WAG/10apr10
Texas A&M 23
What technologies are needed to do this?
WAG/9oct08
Taiwan 24
Local Intelligence
…the physics of emerging technology didn’t work …[using
centralized information systems] … so it is far more effective to
put whatever computing power is required where the data are
located. Efficiency considerations thus favor the distribution of
technology, rather than the concentration of technology. The
economics of information technology are the reverse of those of
mechanical technology.
C. A. Mead
California Institute of Technology
WAG/10apr10
Texas A&M 25
Multi-Agent System
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
WAG/10apr10
A
Agent: an autonomous entity
Attempts to satisfy its local
objectives with independent
actions
Can be functionally
independent of other agents
May be competitive
Usually implemented in
software
Texas A&M 26
Holonic System
Holon: self-contained element
capable of functioning
autonomously in a
cooperative environment
Enables collaboration among
local tasks to achieve a global
objective
Consists of an information
processing part and often a
physical processing part
Can form part of other holons
(“whole-part” relationship)
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
Arthur Koestler, The Ghost in the Machine, Arkana Books,1967
WAG/10apr10
Texas A&M 27
Communication
for peer-to-peer networks
WAG/10apr10
Texas A&M 28
Cooperation
with different objectives …
WAG/10apr10
Texas A&M 29
Collaboration
with a global objective …
WAG/10apr10
Texas A&M 30
Coordination
… based on negotiation protocols
WAG/10apr10
Texas A&M 31
Holonic Technology Platform
Processor
– Processes information gathered by sensors
(RFID, cameras, biometrics, motion)
Memory
– Stores information, applications, and system
software at each node
Transceiver
– Establishes wireless communications with
other nodes
Systems Software
– Intelligent routing of data
– Local decision support
– Distributed processing
WAG/10apr10
Texas A&M 32
System Architecture
C. Ng, Z. Alibhai, D. Sabaz, O. Uncu, and W. A. Gruver, “Framework for developing
distributed systems in a peer-to-peer environment,” Proc. of the IEEE International
Conference on Systems, Man, and Cybernetics, Taipei, Taiwan, October 2006
WAG/10apr10
Texas A&M 33
System Software

User applications developed in an agent framework
System services developed in a holonic framework
Implemented in Java running under Ubuntu Linux O/S
Uses UDP/IP to provide a message based infrastructure for
devices to interconnect
Services include send and receive messages, event-to-event
triggering, virtual network topologies, yellow-pages, remote
agent/holon monitoring and configuration.
Holons facilitate system objectives including communication
routing, where agents should reside for service delivery, and
system security.
WAG/10apr10
Texas A&M 34
Prototype Hardware: 2nd Generation
Processors: Intel PXA270 (640 MHz)
Cirrus ARM920T(200 Mhz)
FPGA: Altera 8256 LUT Cyclone II
Memory: 512Mb SDRAM, 1Gb NOR,
512Mb NAND flash (customizable)
Wireless Transceiver: IEEE 802.11b/g
2.402 - 2.497 GHz, 3 antennas,
100-300 meter range
S. Ovcharenko, Z. Alibhai, C. Ng,
External Interfaces: USB 2.0, RS232, VGA W. A. Gruver, and D. Sabaz,
“Implementation of a wireless
distributed intelligent system,”
Operating System: Debian Linux
Proc. of the 2006 IEEE International
Power Requirements: 5 VDC @ 1.4 A
Workshop on Intelligent Distributed
Systems, Prague, Czech Republic,
Dimensions: 8" x 6" x 2“
June 2006
WAG/10apr10
Texas A&M 35
Prototype System:
3rd Generation
Xilinx FPGA
Virtex 5
UAR
T
Ctrl
High-level software controls
partially reconfigurable user
modules via API
Linux kernel
Xilinx Virtex 5 Dev board,
MicroBlaze Soft Processor
Edward Chen and Victor Gusev,
PhD/MASc students in iDEA Lab
WAG/10apr10
DDR
2
Ctrl
…
ENET
Ctrl
PR
M
1
PR
M
2
ICAP
PR
M
3
User Module
MicroBlaze CPU
High‐Level User App
PR API
Driver
Linux Kernel
PR Manager
Texas A&M 36
Dynamic Partial Reconfiguration
Partially reconfigurable FPGA enables dynamic
reconfiguration without shut down
High-level PR hardware abstraction allows easier
management from user space
Linux provides standard facilities for networking, device
management, etc.
Reduces product cost
Reduces footprint
Reduces power consumption
Increase performance
Faster configuration time
WAG/10apr10
Texas A&M 37
Holonic Intelligence Node
RFID Tag
RFID Antenna
HTP Platform
Digital Signage
RFID Reader
Processor
RFID Tag
Memory
Alarm
Sensors
Transceiver
Holonic Intelligence
Entry Lock
WAG/10apr10
38
Texas A&M 38
Holonic Intelligence Network
WAG/10apr10
Texas A&M 39
HMS Project (1995-2004)
5 Regions
Australia, Canada, European Union, Japan, USA
40 Organizations
Industry
Intelligent Robotics, DaimlerChrysler, Fanuc, GM Holden,
Hitachi, Rockwell Automation, Toshiba, Yaskawa Electric, BHP
Billiton, ANAYAK, ATOS, ATS Spartec, Blastman Robotics,
Okuma, Softing
R&D Labs
Fraunhofer IPA, NRC Canada, CSIRO, Profactor, Tekniker, VTT
Universities
Calgary, Connecticut, Hannover, Kagawa, Keele, Keio, Kobe, KU
Leuven, Osaka Pref., Simon Fraser, Tokyo, Tsukuba, Vanderbilt
WAG/10apr10
Texas A&M 40
Defect Sensitive Manufacturing
Chop Saw
Scanner
Rip Saw
LUMBER
Scanner
Lumber
WAREHOUSE
WAG/10apr10
Texas A&M 41
Holonic Decision Support
Rip each piece of
lumber into a best
combination of
strip by width
Agent
Agent11
Agent
Agent22
Select a best load
or jag of lumber
from the warehouse
Mediator
MediatorAgent
Agent
Agent
Agent33
Optimize component schedule to
produce a best combination of
components for each strip
WAG/10apr10
Texas A&M 42
Rough Mill Decision Support System
Human
Machine
Interface
3D
Virtual
Rough Mill
Reports
Current
Database
Historical
Database
Holonic
Intelligent
Systems
E. Elghoneimy, O. Uncu, W. A. Gruver, and D. B. Kotak, “Simulation and
Decision Support Models for Rough Mills: A Multi-Agent Perspective,”
Proceedings of the 2005 IEEE International Conference on Systems, Man,
and Cybernetics, Hawaii, USA, October 2005
WAG/10apr10
Texas A&M 43
Electric Motor Assembly
Highly variable, small
volume production
Effective interaction
between humans and
industrial robots
Human workers are
essential
Controlled as a holonic
system
HI Display
Parts
pallet
Human Worker
AGV
Vision
Camera
Vision
controller
Tool Pallet
Work
Conveyor
Plug and Production
Linear Slider
New Robot
Yaskawa Electric Company,
HMS Consortium, 2004
WAG/10apr10
Tool Pallet
Cell Controller
Parts pallet
FA LAN
Texas A&M 44
Automated Shot Blasting
Increases efficiency of
automated surface
treatment
Accommodates wide
variety and large sizes of
workpieces
Four gantry robots
24 simultaneous axes
Controlled as a holonic
system
VTT Automation and Blastman Robotics Ltd, HMS Consortium, 1995-2004
WAG/10apr10
Texas A&M 45
Automotive Engine Assembly
DaimlerChrysler plant in Stuttgart, Germany
V6 and V8 engines
USA / Europe / Asia (90 variations)
Assembly of large, heavy engines
Daimler, Fraunhofer IPA, HMS Consortium, 2000-04
WAG/10apr10
Texas A&M 46
Smart Grid
Source: “You think you’re so smart grid,” VTS Enviro Group, May 19, 2009
WAG/10apr10
Texas A&M 47
Automated Meter Reading
Remotely monitor usage
of electricity, water, and
gas
Send data on demand to
utility company for
monitoring and billing
Enable different prices to
be billed for energy
consumption depending
on time of day
Other Appliances
?
?
Gas Meter
WAG/10apr10
Water Meter
To
Next
Router
Electricity
Meter
Texas A&M 48
Smart Home
Source: S. Ahson and M. Ilyas, RFID Handbook, CRC Press, 2008
WAG/10apr10
Texas A&M 49
Integrated Digital Services
Weak Integration
WAG/10apr10
Strong Integration
Texas A&M 50
Conclusions
Holonic intelligence has broad applications to
-
manufacturing and supply chains
energy management
aerospace and defense systems
smart homes
Holonic intelligent systems provide
-
WAG/10apr10
improved flexibility
reduced setup time
higher robustness
improved scalability
integration of human intelligence
Texas A&M 51
Conclusions
Holonic intelligent systems have demonstrated
capabilities to control physical equipment
Holonic intelligent systems offer a migration path from
centralized legacy systems to fully distributed
systems
International standards are being developed for
holonic intelligent systems
WAG/10apr10
Texas A&M 52
Paradigm Change
“The formulation of a problem is often more essential
than its solution, which may be merely a matter of
mathematical or experimental skill. To raise new
questions, new possibilities, to regard old problems
from a new angle, requires creative imagination and
marks real advance in science.”
A. Einstein and L. Infeld
WAG/10apr10
Texas A&M 53
Acknowledgements
North Vancouver, Canada
Intelligent / Distributed Enterprise Automation Laboratory
Simon Fraser University, Burnaby, Canada
Holonic Manufacturing Systems Consortium
Intelligent Manufacturing Systems Program
Distributed Intelligent Systems Technical Committee
IEEE Systems, Man, and Cybernetics Society
WAG/10apr10
Texas A&M 54
For further information …
www.iroboticscorp.com
Intelligent/Distributed Enterprise Automation Laboratory
www.ensc.sfu.ca/idea – select “Publications”
Holonic Manufacturing Consortium
www.ims.org – and select “Completed Projects”
IEEE Technical Committee on Distributed Intelligent
Systems
www.ieeesmc.org - select “Technical Committees”
WAG/10apr10
Texas A&M 55